Epidemiology and virulence-associated genes of Clostridioides difficile isolates and factors associated with toxin EIA results at a university hospital in Japan

Introduction Clostridioides difficile is one of the most important nosocomial pathogens; however, reports regarding its clinical and molecular characteristics from Japan are scarce. Aims We studied the multilocus sequence typing (MLST)-based epidemiology and virulence-associated genes of isolates and the clinical backgrounds of patients from whom the isolates had been recovered. Methods A total of 105 stool samples tested in a C. difficile toxin enzyme immune assay (EIA) were analysed at the University of Tokyo Hospital from March 2013 to July 2014. PCR for MLST and the virulence-associated genes tcdA, tcdB, cdtA, cdtB and tcdC was performed on C. difficile isolates meeting our inclusion criteria following retrospective review of medical records. EIA-positive and EIA-negative groups with toxigenic strains underwent clinical and molecular background comparison. Results The toxigenic strains ST17, ST81, ST2, ST54, ST8, ST3, ST37 and ST53 and the non-toxigenic strains ST109, ST15 and ST100 were frequently recovered. The prevalence rate of tcdA-negative ST81 and ST37, endemic in China and Korea, was higher (11.4%) than that reported in North America and Europe, and hypervirulent ST1(RT027) and ST11(RT078) strains that occur in North America and Europe were not recovered. The linkage between the EIA results and cdt A/B positivity, tcdC deletion, or tcdA variation was absent among toxigenic strains. Compared with the 38 EIA-negative cases, the 36 EIA-positive cases showed that the patients in EIA-positive cases were older and more frequently had chronic kidney disease, as well as a history of beta-lactam use and proton pump inhibitor therapy. Conclusion In Japan, the prevalence rates for tcdA-negative strains are high, whereas the cdtA/B-positive strains are rare. EIA positivity is linked to older age, chronic kidney disease and the use of beta-lactams and proton pump inhibitors.

genomic region [4][5][6]. Pathogenicity is generally defined by the presence of both tcdA and tcdB genes, but there are some exceptions to this relationship because pathogenicity has also been reported for strains with tcdA deletion and those producing only TcdB or CDT [7][8][9]. Other toxin-related genes, such as tcdC, tcdE and tcdR in the PaLoc region, are also thought to be involved in the regulation of toxin A and toxin B production, and the coexistence of deletions in the tcdC and binary toxin productivity has been reported [10][11][12][13][14]. Since the emergence of hypervirulent strains such as NAP1/ RT027 and RT078 producing CDT, the generation of CDT has been strongly associated with increased morbidity, mortality and recurrence rates [13,[15][16][17]. The circulation of strains with various levels of toxigenicity has been reported worldwide, and trends differ between geographical regions [18][19][20].
In microbiological epidemiology studies, PCR ribotyping has historically been the more widely used method, but multilocus sequence typing (MLST) has increasingly been recognized as an equally useful molecular typing method with some advantages over PCR ribotyping in terms of ease of interpretation and the lower interlaboratory variability of test results [21]. Kuwata et al. performed the first microbiological epidemiology study using MLST. However, they did not examine the link between microbiological epidemiology and clinical information [22].
On the basis of this background and with the aim of adding to the few reports on the microbiological epidemiology of C. difficile in Japan, we investigated the toxigenicity and MLSTbased epidemiology of clinical C. difficile isolates and analysed the clinical and microbiological backgrounds by comparing C. difficile toxin enzyme immunoassay (EIA)-positive and EIA-negative cases in a leading university hospital in Japan.

Study design
Molecular genetic analysis of C. difficile isolates recovered from stool samples and tested in the EIA between March 2013 and July 2014 at the University of Tokyo Hospital was conducted. In addition, we performed a retrospective analysis of clinical information in electronic medical records of the patients from whom C. difficile isolates were collected. C. difficile from a previous positive culture in the same diarrhoeal episode and patients without a confirmed record of a diarrhoeal episode or EIA result were excluded. Isolates from patients younger than 18 years old were also excluded due to the possibility of adult and paediatric CDIs having different characteristics [23]. This study was approved by the institutional Ethics Committee.

Isolation of C. difficile strains
During the study period, stool samples submitted to the microbiology laboratory for EIA (C. Diff Quik Chek Complete, TechLab, Inc., Blacksburg, VA, USA) were also anaerobically cultured on cycloserine-cefoxitin-mannitol agar plates (Nissui Pharmaceutical, Co., Ltd, Tokyo, Japan) without any enrichment to enhance bacterial yield for the isolation of C. difficile. Anaerobic culturing was performed at 37 °C for >24 h. C. difficile isolates were identified by colony morphology on agar media and later confirmed by the assessment of the expression of several housekeeping genes as part of the MLST process. Colonies on agar media were routinely inoculated into skim milk and stored at −80 °C.
Molecular assessment of C. difficile isolates C. difficile colonies on culture agar were inoculated into sterile water and then boiled at 95 °C for 10 min to make DNA templates for subsequent PCR with an Emerald Amp PCR Master Mix kit (Takara Bio, Shiga, Japan). MLST and toxin gene analyses (tcdA, tcdB, tcdC, cdtA, cdtB) were performed using primers described in previous reports (Table 1) [22,[24][25][26]. As part of the MLST and tcdC analyses, the PCR amplicon underwent DNA sequencing, and the genomic sequence data of the PaLoc region of the C. difficile strain VPI 10463 (GenBank accession number: X92982.1) was used as reference data to detect the presence of tcdC deletion. Sequence types (STs) were determined based on DNA sequencing data using the PubMLST sequence query page (https:// pubmlst. org/ C. difficile/).

definition and analyses of clinical cases
The medical records of all the patients tested for the toxin by the EIA test were reviewed. Patients without episodes of unformed stool on electronic clinical records and those younger than 18 years old at the time of testing were excluded. Cases whose anaerobic stool cultures were negative for C. difficile were also excluded, and available C. difficile isolates were used for further molecular analysis. The clinical parameters retrieved for analysis in our study were age, sex, dates of admission and discharge if hospitalized, clinical outcome including death, length of hospitalization at EIA testing, antibiotic administration within 1 month prior to stool testing, presence of comorbidities such as diabetes mellitus, chronic kidney disease, malignancy, chemotherapy within 1 month, neutropaenia (<500 μl −1 ), daily use of immunosuppressive agents within 1 week, history of abdominal surgery, tube feeding status, and daily use of probiotics and proton pump inhibitors. White blood cell counts and serum creatinine [on the day EIA stool sample submitted if available; if not, on the closest (≤2) day after the EIA test], as well as vital signs, were also reviewed to score severity based on the definition of severity in the initial episodes of C. difficile infection shown in the Infectious Diseases Society of America (IDSA) clinical practice guidelines for CDI management, updated in 2017: cases with hypotension, shock, ileus, megacolon were labelled as fulminant cases, and cases with white blood cell counts ≥15,000 µl −1 or serum creatinine levels >1.5 mg dl −1 were labelled as severe cases [27]. Cases not labelled as fulminant or severe were categorized as non-severe. Clinical information for the EIA-positive and EIA-negative groups was compared, and the clinical backgrounds, outcomes and toxigenicity of the C. difficile isolates from the EIA-positive group were integrated with clinical information and underwent statistical analysis.

Statistical analysis
Statistical analysis was performed using JMP software (JMP Pro 14, SAS Institute Japan, Tokyo, Japan). For categorical data and discrete measurement data analyses, the chi-squared test and Wilcoxon test were used, respectively. The Kaplan-Meier estimate was tested using the log-rank test to analyse potential differences in mortality rates between the groups.
All statistical analyses were conducted with a significance level of α=0.05 (P<0.05).

Inclusion criteria of C. difficile isolates and clinical cases
A total of 105 C. difficile isolates from the same number of clinical cases were included in this study after our inclusion criteria had been applied. The 105 isolates underwent MLST Table 1. List of primers used in this study target

Microbiological epidemiology of C. difficile isolates
Information about ST distribution, identified clades and toxin gene presence is shown in Table 2. ST17 (n=12), ST81 (n=9), ST2ㆍST15ㆍST54ㆍST109 (n=7) and ST8ㆍST3ㆍST100 (n=4) were the most prevalent STs. Clade analysis that is commonly used to view clustering of STs with similar characteristics revealed that clade 1 strains were the most prevalent (n=72), followed by those of clade 4 (n=24). Fewer strains belonged to clades 2 and 3, and some strains could not be categorized by the data available on PubMLST. Major STs belonging to each clade are highlighted in Table 3. The proportions of isolates from EIA-positive and EIA-negative cases are also shown for commonly isolated STs in Table 4.

Relationships between toxigenicity, cdtA/B, tcdA variation and tcdC deletion with ST
As shown in Table 2, out of the 105 isolates studied, 74 were toxigenic and 31 were non-toxigenic. All toxigenic strains were confirmed to possess the tcdC gene, whereas nontoxigenic strains did not express it. The presence of cdtA/ cdtB genes, tcdA deletions and tcdC deletions was confirmed in 5, 12 and 6 isolates, respectively. Of the six isolates with the tcdC deletion, five were cdtA/cdtB-positive, whereas one was tcdA+/tcdB+ and cdtA/B-negative. Except for ST3, no discrepancy in toxin gene patterns was observed between isolates in the same ST. In ST3, three isolates were toxigenic strains, whereas one isolate was non-toxigenic.

Toxin EIA test results and tcdC deletion, binary toxin and toxin A variation
A total of 74 clinical cases were confirmed to be toxigenic, and the toxin EIA test was positive in 36 cases and negative in 38 cases. The most frequently found strains in the EIApositive and EIA-negative groups are listed in Table 4. No significant differences between the two groups regarding the prevalence rates of tcdA-negative variants, cdtA/B-positive   strains and strains with tcdC deletion in the EIA-positive and EIA-negative groups were found ( Table 5).

Comparison of clinical background between EIApositive and EIA-negative patients
Of The other clinical factors analysed in our study did not differ significantly between these groups. A complete list of the clinical factors analysed in this study is provided in Table 6.

Microbiological epidemiology and virulenceassociated genes
This study is one of the most thorough microbiological analyses of the epidemiology and virulence factors of C. difficile isolates in Japan, and also provides an insight into the clinical background of cases from which C. difficile were isolated through comparisons based on toxin presence revealed by EIA.
Epidemiological analysis of the C. difficile isolates in our study is essential as MLST-based reports from Japan are scarce.
The relatively high prevalence rates for tcdA-negative/tcdBpositive strains ST81 and ST37 in our study is in the range of previously reported prevalence rates of 10-30 % in Asian countries, including China and Korea [31][32][33][34], and higher than that reported in non-Asian regions, where only rare outbreaks of tcdA-negative/tcdB-positive strains have been reported [35,36]. In the USA and Europe, the fractions of tcdA-negative/tcdB-positive strains recovered from clinical stool samples were reported to be less than 3-5 %, respectively [19,37].
The rarity of binary toxin-producing strains, including ST1(RT027) or ST11(RT078) in our study, is consistent with reports from Japan's neighbouring countries showing the same tendency compared with North America or Europe [22,[31][32][33][34]. The binary toxin-producing ST41, ST123 and ST223 strains in our study are also reported to cause severe CDI [25,38]. In previous reports, it was also shown that these strains share cdtA/B, 18 bp deletion in tcdC and fluoroquinolone resistance with the ST1/RT027 strain in the same clade (2) [25,[38][39][40], but differences between these strains and the ST1/RT027 strain, such as the structure and cytotoxicity of TcdB and toxin productivity, have also been reported [38,41]. An outbreak of these strains may have a significant impact in countries where ST1/RT027 strains are rarely found, and further insight into the pathogenicity of these strains is warranted.
With respect to non-toxigenic strains, the relatively high prevalence rate of non-toxigenic strains ST15, ST109 and ST100 was consistent with a previous study performed in Japan [22]. Non-toxigenic strains are of interest because there are reports concerning the protective effect of colonization by non-toxigenic strains against CDI [42], and some preclinical and clinical studies have shown that the administration of certain selected non-toxigenic C. difficile strains could play a defensive role against CDI development [43][44][45][46]. Our epidemiological data on non-toxigenic strains may contribute to future treatment or prevention measures against CDI.
The discrepancy in toxigenicity is known to be observed among isolates in the same strain, and the discrepancy in toxigenicity in ST3 in our study was also shown previously [33]. Other strains, including ST109 and ST100, have also been shown to encompass both toxigenic and non-toxigenic strains [47,48], and this issue warrants further evolutionary study [49].

Comparison of EIA-positive and EIA-negative groups
Comparison between EIA-positive and EIA-negative cases was performed as in daily practice. Our hospital and many other hospitals utilize EIA test results for CDI diagnosis and management. Previous studies have shown that EIA positivity is positively linked to the presence of cdtA/B [50,51] and tcdC deletion [52]. However, given the limited statistical power of our study, we could only suggest that cdtA/B positivity correlates with EIA positivity. The proportion of tcdA-negative isolates was not significantly different between the two groups, which is consistent with previous findings [33,53].
The comparison between the EIA-positive and EIA-negative groups highlighted advanced age, hospital stay of longer than 3 days, chronic kidney disease, and use of beta-lactam antibiotics and proton pump inhibitors as factors that correlated with the positive EIA result. These factors are reportedly recognized as critical elements that increase the risk of CDI development [54][55][56][57][58]. Although seemingly contradicting the results of two previous studies [59,60], in our study, the history of probiotic use was not linked to a higher rate of EIA-positive results.

Limitation
The small number of clinical cases (partly due to the exclusion of paediatric cases) and the limited availability of C. difficile isolates due to the nature of our single institution-based study are the limitations of our study.

Conclusion
Our study makes a significant contribution to the field as one of the first MLST-based epidemiological studies performed in Japan. The relatively high prevalence of toxin A-negative/B-positive strains and the low prevalence of hypervirulent binary toxin-producing strains observed in this study were similar to the epidemiological data reported in previous studies in Japan. Older age, recent use of beta-lactam antibiotics and proton pump inhibitors, and more prolonged hospitalization in the EIA-positive group were demonstrated in our study, whereas no correlations between EIA positivity and microbiological virulence-associated factors were found.

Funding information
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.